Temporal Instability of Commuting Determinants Evidence from Electric Vehicle Telematics Data in South Korea

This study highlights the potential risks of assuming coefficient invariance in annual cross-sectional data, particularly in the context of high-frequency behavioral shifts. Using a one-year longitudinal dataset of electric vehicle travel logs, this study employs a state-space model to estimate time-varying parameters. The results reveal a clear dichotomy: structural factors (departure time, location) remain stable, while exogenous factors (weather, battery status) exhibit high behavioral volatility. The significant state variance in the intercept and exogenous variables suggests that static analyses run the risk of misinterpreting transient noise as structural relationships for specific exogenous factors. These findings indicate that commuting determinants are not constant but time-varying, suggesting that dynamic longitudinal approaches can complement static analyses by revealing temporal variations that single-point estimates may overlook.